[PATCH v9 5/8] PCI: imx6: Refine the regulator usage

[Lucas Stach]

Am Donnerstag, dem 09.06.2022 um 06:17 +0000 schrieb Hongxing Zhu:
> > -----Original Message-----
> > From: Lucas Stach <l.stach at pengutronix.de>
> > Sent: 2022年6月8日 15:27
> > To: Hongxing Zhu <hongxing.zhu at nxp.com>; bhelgaas at google.com;
> > robh+dt at kernel.org; broonie at kernel.org; lorenzo.pieralisi at arm.com;
> > jingoohan1 at gmail.com; festevam at gmail.com;
> > francesco.dolcini at toradex.com
> > Cc: linux-pci at vger.kernel.org;
> > linux-arm-kernel at lists.infradead.org;
> > linux-kernel at vger.kernel.org; kernel at pengutronix.de; dl-linux-imx
> > <linux-imx at nxp.com>
> > Subject: Re: [PATCH v9 5/8] PCI: imx6: Refine the regulator usage
> > 
> > Am Freitag, dem 06.05.2022 um 09:47 +0800 schrieb Richard Zhu:
> > > The driver should undo any enables it did itself. The regulator
> > > disable shouldn't be basing decisions on regulator_is_enabled().
> > > 
> > > To keep the balance of the regulator usage counter, disable the
> > > regulator just behind of imx6_pcie_assert_core_reset() in resume
> > > and
> > shutdown.
> > > 
> > > Signed-off-by: Richard Zhu <hongxing.zhu at nxp.com>
> > > ---
> > >  drivers/pci/controller/dwc/pci-imx6.c | 19 +++++++------------
> > >  1 file changed, 7 insertions(+), 12 deletions(-)
> > > 
> > > diff --git a/drivers/pci/controller/dwc/pci-imx6.c
> > > b/drivers/pci/controller/dwc/pci-imx6.c
> > > index 7005a7910003..3ce3993d5797 100644
> > > --- a/drivers/pci/controller/dwc/pci-imx6.c
> > > +++ b/drivers/pci/controller/dwc/pci-imx6.c
> > > @@ -369,8 +369,6 @@ static int imx6_pcie_attach_pd(struct device
> > > *dev)
> > > 
> > >  static void imx6_pcie_assert_core_reset(struct imx6_pcie
> > > *imx6_pcie)
> > > {
> > > -	struct device *dev = imx6_pcie->pci->dev;
> > > -
> > >  	switch (imx6_pcie->drvdata->variant) {
> > >  	case IMX7D:
> > >  	case IMX8MQ:
> > > @@ -400,14 +398,6 @@ static void
> > > imx6_pcie_assert_core_reset(struct
> > imx6_pcie *imx6_pcie)
> > >  				   IMX6Q_GPR1_PCIE_REF_CLK_EN, 0
> > > << 16);
> > >  		break;
> > >  	}
> > > -
> > > -	if (imx6_pcie->vpcie && regulator_is_enabled(imx6_pcie-
> > > >vpcie) > 0) {
> > > -		int ret = regulator_disable(imx6_pcie->vpcie);
> > > -
> > > -		if (ret)
> > > -			dev_err(dev, "failed to disable vpcie
> > > regulator: %d\n",
> > > -				ret);
> > > -	}
> > >  }
> > > 
> > >  static unsigned int imx6_pcie_grp_offset(const struct imx6_pcie
> > > *imx6_pcie) @@ -580,7 +570,7 @@ static int
> > imx6_pcie_deassert_core_reset(struct imx6_pcie *imx6_pcie)
> > >  	struct device *dev = pci->dev;
> > >  	int ret, err;
> > > 
> > > -	if (imx6_pcie->vpcie && !regulator_is_enabled(imx6_pcie-
> > > >vpcie)) {
> > > +	if (imx6_pcie->vpcie) {
> > >  		ret = regulator_enable(imx6_pcie->vpcie);
> > >  		if (ret) {
> > >  			dev_err(dev, "failed to enable vpcie
> > > regulator: %d\n", @@
> > -653,7
> > > +643,7 @@ static int imx6_pcie_deassert_core_reset(struct
> > > imx6_pcie
> > *imx6_pcie)
> > >  	return 0;
> > > 
> > >  err_clks:
> > > -	if (imx6_pcie->vpcie && regulator_is_enabled(imx6_pcie-
> > > >vpcie) > 0) {
> > > +	if (imx6_pcie->vpcie) {
> > >  		ret = regulator_disable(imx6_pcie->vpcie);
> > >  		if (ret)
> > >  			dev_err(dev, "failed to disable vpcie
> > > regulator: %d\n", @@
> > -1026,6
> > > +1016,9 @@ static int imx6_pcie_resume_noirq(struct device *dev)
> > >  		return 0;
> > > 
> > >  	imx6_pcie_assert_core_reset(imx6_pcie);
> > > +	if (imx6_pcie->vpcie)
> > > +		regulator_disable(imx6_pcie->vpcie);
> > > +
> > This one looks misplaced. Surely you want the regulator to be on
> > when
> > resuming the PCIe subsystem. Isn't this just papering over a wrong
> > usage count
> > here, because there is no regulator_disable in
> > imx6_pcie_suspend_noirq,
> > where I would expect this to happen?
> > 
> Hi Lucas:
> Thanks for your comments.
> There was one regulator_disable() operation at the end of the
>  imx6_pcie_assert_core_reset() function before.
> When create the 5/8 patch, I follow the same behavior to disable the
> regulator just behind the imx6_pcie_assert_core_reset() function.
> 
> Yes, it is. Imx6_pcie_suspend_noirq doesn't have the
> regulator_disable.
> The regulaor_enable is contained in imx6_pcie_deassert_core_reset().
> Both of the regulator_disable and regulator_enabe are invoked once in
>  imx6_pcie_resume_noirq.
> So, the regulator is on and has a balanced usage count after resume.
> 

Yea, my argument is that when we are moving around the regulator
handling anyways, we should move the regulator_disable into the suspend
function. It's the right thing to do: we don't want the bus to be
powered when the system is in suspend and while the use-count is
correct, it's confusing to read the resume function otherwise.

Regards,
Lucas